Mean Sensitivity Values Research Articles

AutoScore-Imbalance: An interpretable machine learning tool for development of clinical scores with rare events data

BackgroundMedical decision-making impacts both individual and public health. Clinical scores are commonly used among various decision-making models to determine the degree of disease deterioration at the bedside. AutoScore was proposed as a useful clinical score generator based on machine learning and a generalized linear model. However, its current framework still leaves room for improvement when addressing unbalanced data of rare events. MethodsUsing machine intelligence approaches, we developed AutoScore-Imbalance, which comprises three components: training dataset optimization, sample weight optimization, and adjusted AutoScore. Baseline techniques for performance comparison included the original AutoScore, full logistic regression, stepwise logistic regression, least absolute shrinkage and selection operator (LASSO), full random forest, and random forest with a reduced number of variables. These models were evaluated based on their area under the curve (AUC) in the receiver operating characteristic analysis and balanced accuracy (i.e., mean value of sensitivity and specificity). By utilizing a publicly accessible dataset from Beth Israel Deaconess Medical Center, we assessed the proposed model and baseline approaches to predict inpatient mortality. ResultsAutoScore-Imbalance outperformed baselines in terms of AUC and balanced accuracy. The nine-variable AutoScore-Imbalance sub-model achieved the highest AUC of 0.786 (0.732–0.839), while the eleven-variable original AutoScore obtained an AUC of 0.723 (0.663–0.783), and the logistic regression with 21 variables obtained an AUC of 0.743 (0.685–0.801). The AutoScore-Imbalance sub-model (using a down-sampling algorithm) yielded an AUC of 0.771 (0.718–0.823) with only five variables, demonstrating a good balance between performance and variable sparsity. Furthermore, AutoScore-Imbalance obtained the highest balanced accuracy of 0.757 (0.702–0.805), compared to 0.698 (0.643–0.753) by the original AutoScore and the maximum of 0.720 (0.664–0.769) by other baseline models. ConclusionsWe have developed an interpretable tool to handle clinical data imbalance, presented its structure, and demonstrated its superiority over baselines. The AutoScore-Imbalance tool can be applied to highly unbalanced datasets to gain further insight into rare medical events and facilitate real-world clinical decision-making.

Utility of the Simulated Outcomes Following Carotid Artery Laceration Video Data Set for Machine Learning Applications

Surgical data scientists lack video data sets that depict adverse events, which may affect model generalizability and introduce bias. Hemorrhage may be particularly challenging for computer vision-based models because blood obscures the scene. To assess the utility of the Simulated Outcomes Following Carotid Artery Laceration (SOCAL)-a publicly available surgical video data set of hemorrhage complication management with instrument annotations and task outcomes-to provide benchmarks for surgical data science techniques, including computer vision instrument detection, instrument use metrics and outcome associations, and validation of a SOCAL-trained neural network using real operative video. For this quailty improvement study, a total of 75 surgeons with 1 to 30 years' experience (mean, 7 years) were filmed from January 1, 2017, to December 31, 2020, managing catastrophic surgical hemorrhage in a high-fidelity cadaveric training exercise at nationwide training courses. Videos were annotated from January 1 to June 30, 2021. Surgeons received expert coaching between 2 trials. Hemostasis within 5 minutes (task success, dichotomous), time to hemostasis (in seconds), and blood loss (in milliliters) were recorded. Deep neural networks (DNNs) were trained to detect surgical instruments in view. Model performance was measured using mean average precision (mAP), sensitivity, and positive predictive value. SOCAL contains 31 443 frames with 65 071 surgical instrument annotations from 147 trials with associated surgeon demographic characteristics, time to hemostasis, and recorded blood loss for each trial. Computer vision-based instrument detection methods using DNNs trained on SOCAL achieved a mAP of 0.67 overall and 0.91 for the most common surgical instrument (suction). Hemorrhage control challenges standard object detectors: detection of some surgical instruments remained poor (mAP, 0.25). On real intraoperative video, the model achieved a sensitivity of 0.77 and a positive predictive value of 0.96. Instrument use metrics derived from the SOCAL video were significantly associated with performance (blood loss). Hemorrhage control is a high-stakes adverse event that poses unique challenges for video analysis, but no data sets of hemorrhage control exist. The use of SOCAL, the first data set to depict hemorrhage control, allows the benchmarking of data science applications, including object detection, performance metric development, and identification of metrics associated with outcomes. In the future, SOCAL may be used to build and validate surgical data science models.

Accuracy of MRI, CT, and Ultrasound imaging on thickness and depth of oral primary carcinomas invasion: a systematic review.

The aim of this study was to perform a systematic review to assess the sensitivity, specificity, and accuracy of magnetic resonance imaging (MRI), computed tomography (CT), and intraoral ultrasound (US) to determine the depth of invasion (DOI) and/or tumor thickness (TT) in oral cavity cancers, with histopathological evaluation as the gold standard. Articles whose primary objective was to evaluate the accuracy (sensitivity and specificity) of MRI, CT, and US imaging to assess DOI and/or TT were searched in six major electronic databases, in addition to three grey literature databases. The methodological quality of the selected studies was evaluated by using the Quality Assessment Tool for Diagnostic Accuracy Studies-2. Twelve studies met the inclusion criteria and underwent qualitative analysis: six studies on MRI, three on US, and one on CT. The accuracy values for MRI ranged from 67 to 83%, with sensitivity values above 80% and specificity above 75%. For US, mean values of sensitivity ranged from 91 to 93%. For CT, accuracy was 75%. For the application of US, CT, MRI, good accuracy was reported in DOI and/or TT, as evaluated in the preoperative period. US offered advantages for detection of small lesions.

Structure-Function Relationships and Glaucoma Detection with Magnification Correction of OCT Angiography.

PurposeTo investigate the effects of adjusting the ocular magnification during OCT-based angiography imaging on structure–function relationships and glaucoma detection.DesignCross-sectional study.ParticipantsA total of 96 healthy control participants and 90 patients with open-angle glaucoma were included.MethodsOne eye of each patient in the control group and the patient group was evaluated. The layers comprising the macula vascular density (VD) and circumpapillary VD were derived from swept-source OCT angiography imaging. The mean sensitivity (MS) of the standard automated perimetry was measured using the Humphrey 24-2 test. Structure–function relationships were evaluated with simple and partial correlation coefficients. A receiver operating characteristic analysis was performed to evaluate the diagnostic accuracy for glaucoma using the area under the receiver operating characteristic curve (AUC). Ocular magnification was adjusted using Littmann’s formula modified by Bennett.Main Outcome MeasuresThe association between the axial length and VD, structure–function relationships, and glaucoma detection with and without magnification correction.ResultsThe superficial layer of the macular region was not significantly correlated to the axial length without magnification correction (r = 0.0011; P = 0.99); however, it was negatively correlated to the axial length with magnification correction (r = –0.22; P = 0.028). Regarding the nerve head layer in the circumpapillary region, a negative correlation to the axial length without magnification correction was observed (r = –0.22; P = 0.031); however, this significant correlation disappeared with magnification correction. The superficial layer of the macula and the nerve head layer of the circumpapillary region were significantly correlated to Humphrey 24-2 MS values without magnification correction (r = 0.22 and r = 0.32, respectively); however, these correlations did not improve after magnification correction (r = 0.20 and r = 0.33, respectively). Glaucoma diagnostic accuracy in the superficial layer (AUC, 0.63) and nerve head layer (AUC, 0.70) without magnification correction did not improve after magnification correction (AUC, 0.62 and 0.69, respectively).ConclusionsAdjustment of the ocular magnification is important for accurate VD measurements; however, it may not significantly impact structure–function relationships and glaucoma detection.

A Deep Learning Approach to Segment and Classify C-Shaped Canal Morphologies in Mandibular Second Molars Using Cone-beam Computed Tomography

IntroductionThe identification of C-shaped root canal anatomy on radiographic images affects clinical decision making and treatment. The aims of this study were to develop a deep learning (DL) model to classify C-shaped canal anatomy in mandibular second molars from cone-beam computed tomographic (CBCT) volumes and to compare the performance of 3 different architectures. MethodsU-Net, residual U-Net, and Xception U-Net architectures were used for image segmentation and classification of C-shaped anatomies. Model training and validation were performed on 100 of a total of 135 available limited field of view CBCT images containing mandibular molars with C-shaped anatomy. Thirty-five CBCT images were used for testing. Voxel-matching accuracy of the automated labeling of the C-shaped anatomy was assessed with the Dice index. The mean sensitivity of predicting the correct C-shape subcategory was calculated based on detection accuracy. One-way analysis of variance and post hoc Tukey honestly significant difference tests were used for statistical evaluation. ResultsThe mean Dice coefficients were 0.768 ± 0.0349 for Xception U-Net, 0.736 ± 0.0297 for residual U-Net, and 0.660 ± 0.0354 for U-Net on the test data set. The performance of the 3 models was significantly different overall (analysis of variance, P = .000779). Both Xception U-Net (Q = 7.23, P = .00070) and residual U-Net (Q = 5.09, P = .00951) performed significantly better than U-Net (post hoc Tukey honestly significant difference test). The mean sensitivity values were 0.786 ± 0.0378 for Xception U-Net, 0.746 ± 0.0391 for residual U-Net, and 0.720 ± 0.0495 for U-Net. The mean positive predictive values were 77.6% ± 0.1998% for U-Net, 78.2% ± 0.0.1971% for residual U-Net, and 80.0% ± 0.1098% for Xception U-Net. The addition of contrast-limited adaptive histogram equalization had improved overall architecture efficacy by a mean of 4.6% (P < .0001). ConclusionsDL may aid in the detection and classification of C-shaped canal anatomy.

MRI and spondyloarthropathy: diagnostic performance compared to long-term clinical follow-up with evaluation of gadolinium chelates injection.

To evaluate MRI performance on both initial and long-term rheumatologic diagnosis of spondyloarthritis (SpA), taking into account clinical evolution and treatment response, and the impact of gadolinium injection. In this single-center study, patients who underwent both spinal and sacroiliac (SI) joint MRI were prospectively recruited between May 2013 and January 2014 and followed for 7 years until 2020. Clinical, biological, and radiologic parameters were collected. At 7-year follow-up (2020), two independent readers reevaluated the initial MRI datasets for specific radiological features of SpA with a 5-point Likert scale to record the estimation of confidence. The centralized MRI interpretations were compared to the established rheumatologic diagnoses in 2013 and 2020. In total, 145 patients (52 men and 93 women) were included. During the 7-year follow-up, the number of patients with positive SpA diagnosis decreased from 93 to 58. Mean sensitivity, specificity, and accuracy of non-contrast MRI were 18, 97, and 49% and 27, 97, and 69% considering 2013 and 2020 rheumatologic diagnoses, respectively. Mean sensitivity, specificity, and accuracy values of gadolinium-enhanced MRI were 26, 97, and 54% and 38, 97, and 73% considering 2013 and 2020 diagnoses, respectively. Post-contrast MRI enabled identification of a subgroup of enthesis-only lesions, without any bone lesions, corresponding to 14% of the pathological cohort. It confirmed uncertain diagnoses in an additional 8.5% of pathological cases. MRI performance for SpA diagnosis is higher when long-term clinical follow-up is considered than when compared to initial diagnosis. Gadolinium injection increases MRI diagnostic performance and may demonstrate a pure enthesic form of the disease, without bone abnormality. • Compared to the rheumatologist's diagnosis over long-term clinical follow-up, MRI performance for SpA is higher than usually estimated. • Gadolinium injection increases diagnostic performance of MRI as it may identify a purely enthesis form of the disease. • Gadolinium injection should be discussed in patients for whom the diagnostic suspicion is strong and whose initial non-injected examination is normal or doubtful.

Comparison of a full arch digital photographic assessment of caries prevalence in 5-year-old children to an established visual assessment method: a cross-sectional study

IntroductionDigital epidemiology in dental disease screening has a number of advantages which warrant further exploration.AimThis study aimed to test the examination accuracy of digital images to evaluate child oral health by comparing the new method to a gold standard method. It also investigated the levels of diagnostic accuracy between different examiners, including dental care professionals and a lay examiner, when quantifying dental disease using images.MethodsA calibrated dental examiner inspected forty 5-year-olds. In addition, three sets of digital images were taken per child. These images were assessed by six examiners. Sensitivity and specificity of caries diagnosis and inter-examiner reliability were calculated to compare the caries scores derived from examination of the images to those of the gold standard examinations.ResultsThe mean values for sensitivity and specificity scores were 48.0% and 99.1%, respectively. The mean value for kappa showed moderate agreement between 0.43 and 0.73 (0.57). Mean values for agreement using intra-class coefficients were excellent (0.78) and good (0.73) for dt and dmft, respectively. No statistical difference in the validity of the caries scores was shown between the different image assessors.ConclusionsThese data demonstrate the feasibility of using digital images to screen child oral health and for nondental professionals to be recruited to carry out digital epidemiology for the oral health surveillance of children.

Automatic segmentation of uterine endometrial cancer on multi-sequence MRI using a convolutional neural network

Endometrial cancer (EC) is the most common gynecological tumor in developed countries, and preoperative risk stratification is essential for personalized medicine. There have been several radiomics studies for noninvasive risk stratification of EC using MRI. Although tumor segmentation is usually necessary for these studies, manual segmentation is not only labor-intensive but may also be subjective. Therefore, our study aimed to perform the automatic segmentation of EC on MRI with a convolutional neural network. The effect of the input image sequence and batch size on the segmentation performance was also investigated. Of 200 patients with EC, 180 patients were used for training the modified U-net model; 20 patients for testing the segmentation performance and the robustness of automatically extracted radiomics features. Using multi-sequence images and larger batch size was effective for improving segmentation accuracy. The mean Dice similarity coefficient, sensitivity, and positive predictive value of our model for the test set were 0.806, 0.816, and 0.834, respectively. The robustness of automatically extracted first-order and shape-based features was high (median ICC = 0.86 and 0.96, respectively). Other high-order features presented moderate-high robustness (median ICC = 0.57–0.93). Our model could automatically segment EC on MRI and extract radiomics features with high reliability.

A new Protophyllocladoxylon stem from the Xishanyao Formation (Middle Jurassic) in the Santanghu Basin, Xinjiang, Northwest China

A new permineralized stem, Protophyllocladoxylon yiwuense Gou et Feng sp. nov., is described from the Middle Jurassic Xishanyao Formation in Naomaohu Town of Yiwu County, Hami City, Xinjiang Uygur Autonomous Region, Northwest China. The stem is only preserved with secondary xylem, which is pycnoxylic and exclusively composed of tracheids and parenchymatous rays. The tracheids have uniseriate or biseriate, alternately arranged bordered pits on their radial walls, and uniseriate bordered pits on the tangential walls. The rays are uniseriate or partially biseriate, up to 34 cells high. There are one or two large, simple pits in each cross field. A branch trace horizontally passes through the secondary xylem showing anatomical structures identical to the secondary xylem. Analyses of the growth-ring width and mean sensitivity values indicate that P. yiwuense had long growing seasons with readily available water supplies and favorable paleoclimatic conditions. White-rot decay features are commonly recognized in the stem, including pocket-like cavities, complete removal of the middle lamellae, thickened corners and separation of the secondary walls of tracheids, as well as septa-like structures in the tracheid lumens due to the cell's reactions to fungal activity. Our study sheds new light into the plant diversity and paleoenvironmental and paleoecological conditions of the Middle Jurassic.

A sustainable IoHT based computationally intelligent healthcare monitoring system for lung cancer risk detection

A sustainable healthcare focuses on enhancing and restoring public health parameters thereby reducing gloomy impacts on social, economic and environmental elements of a sustainable city. Though it has uplifted public health, yet the rise of chronic diseases is a concern in sustainable cities. In this work, a sustainable lung cancer detection model is developed to integrate the Internet of Health Things (IoHT) and computational intelligence, causing the least harm to the environment. IoHT unit retains connectivity continuously generates data from patients. Heuristic Greedy Best First Search (GBFS) algorithm is used to select most relevant attributes of lung cancer data upon which random forest algorithm is applied to classify and differentiates lung cancer affected patients from normal ones based on detected symptoms. It is observed during the experiment that the GBFS-Random forest model shows a promising outcome. While an optimal accuracy of 98.8 % was generated, simultaneously, the least latency of 1.16 s was noted. Specificity and sensitivity recorded with the proposed model on lung cancer data are 97.5 % and 97.8 %, respectively. The mean accuracy, specificity, sensitivity, and f-score value recorded is 96.96 %, 96.26 %, 96.34 %, and 96.32 %, respectively, over various types of cancer datasets implemented. The developed smart and intelligent model is sustainable. It reduces unnecessary manual overheads, safe, preserves resources and human resources, and assists medical professionals in quick and reliable decision making on lung cancer diagnosis.

Automated Detection and Tortuosity Characterization of Retinal Vascular Networks

Automated retinal vascular network detection and analysis using digital retinal images continue to play a major role in the field of biomedicine for the diagnosis and management of various forms of human ailments like hypertension, diabetic retinopathy, retinopathy of prematurity, glaucoma and cardiovascular diseases. Although several literature have implemented different automatic approaches of detecting blood vessels in the retinal and also determining their tortuous states, the results obtained show that there are needs for further investigation on more efficient ways to detect and characterize the blood vessel network tortuosity states. This paper implements the use of an adaptive thresholding method based on local spatial relational variance (LSRV) for the detection of the retinal vascular networks. The suitability of a multi-layer perceptron artificial neural network (MLP-ANN) technique for the tortuosity characterization of retinal blood vascular networks is also presented in this paper. Some vessel geometric features of detected vessels are fed into ANN classifier for the automatic classification of the retinal vascular networks as being tortuous vessels or normal vessels. Experimental studies conducted on DRIVE and STARE databases show that the vascular network detection results obtained from the method implemented in this paper detects large and thin vascular networks in the retina. In comparison to preious methods in the literature, the proposed method for vascular network segmentation achieved better performance than several methods in the literature with a mean accuracy value of 95.04% and mean sensitivity value of 75.16% on DRIVE and mean accuracy value of 94.02% and average sensitivity value of 76.55% on STARE with computational processing time of 4.5 seconds and 9.4 seconds on DRIVE and STARE respectively. The MLP-ANN method proposed for the vascular network tortuosity characterization achieves promising accuracy rates of 77.5%, 80%, 83.33%, 85%, 86.67% and 100% for varying training sample sizes.

Location and stability of the preferred retinal locus in native Persian‐speaking patients with age‐related macular degeneration

ABSTRACT Clinical relevance The findings of this study can be used in the selection of the preferred retinal locus to establish better rehabilitation services such as eccentric viewing training for patients with age‐related macular degeneration. Background The aim of this study was to determine the characteristics of the preferred retinal locus in native Persian‐speaking patients with age‐related macular degeneration. Methods In this non‐interventional case series, all patients with a diagnosis of age‐related macular degeneration referred to the Retina Clinic of the Rassoul Akram Hospital, Tehran, Iran, were evaluated. The fixation characteristics were evaluated monocularly using the MP1 microperimeter (Nidek Technologies, Padua, Italy). Optical coherence tomography was used to determine the location of the central fovea. The images were overlaid and the preferred retinal locus‐fovea distance was measured using Image J software. Results Fifty‐one eyes of 35 patients with a mean age of 73.8 ± 7.7-years were evaluated in this study. Inferior‐field, left‐field, central‐field, right‐field, and superior‐field preferred retinal locus were detected in 49 per cent, 33.3 per cent, 7.8 per cent, 5.9 per cent, and 3.9 per cent of the subjects, respectively. Fixation was stable in 70.6 per cent, relatively unstable in 15.7 per cent, and unstable in 13.7 per cent of the participants. Significant differences were not found in the mean values of logMAR visual acuity between different fields of the preferred retinal locus after Bonferroni correction (p = 0.031). Analysis of co‐variance showed no significant difference in mean sensitivity values between different locations of the preferred retinal locus (p = 0.07). The mean preferred retinal locus‐fovea distance was not significantly different between different fields of the preferred retinal locus (p = 0.063). Conclusions Native Persian‐speaking patients with central scotoma secondary to age‐related macular degeneration place their self‐selected preferred retinal locus most frequently in the inferior and left visual field, which would result in scotoma displacement to the superior and right visual field. Fixation stability was statistically similar in different locations of preferred retinal locus, but it improved with decreasing the preferred retinal locus‐fovea distance.

Texture feature-based machine learning classifier could assist in the diagnosis of COVID-19

PurposeDifferentiating COVID-19 from other acute infectious pneumonias rapidly is challenging at present. This study aims to improve the diagnosis of COVID-19 using computed tomography (CT). MethodCOVID-19 was confirmed mainly by virus nucleic acid testing and epidemiological history according to WHO interim guidance, while other infectious pneumonias were diagnosed by antigen testing. The texture features were extracted from CT images by two radiologists with 5 years of work experience using modified wavelet transform and matrix computation analyses. The random forest (RF) classifier was applied to identify COVID-19 patients and images. ResultsWe retrospectively analysed the data of 95 individuals (291 images) with COVID-19 and 96 individuals (279 images) with other acute infectious pneumonias, including 50 individuals (160 images) with influenza A/B. In total, 6 texture features showed a positive association with COVID-19, while 4 features were negatively associated. The mean AUROC, accuracy, sensitivity, and specificity values of the 5-fold test sets were 0.800, 0.722, 0.770, and 0.680 for image classification and 0.858, 0.826, 0.809, and 0.842 for individual classification, respectively. The feature ‘Correlation’ contributed most both at the image level and individual level, even compared with the clinical factors. In addition, the texture features could discriminate COVID-19 from influenza A/B, with an AUROC of 0.883 for images and 0.957 for individuals. ConclusionsThe developed texture feature-based RF classifier could assist in the diagnosis of COVID-19, which could be a rapid screening tool in the era of pandemic.

Subscapularis (SSC) tendon tears: diagnostic performance and reliability of magnetic resonance arthrography (MRA) with arthroscopic correlation and comparison with clinical tests

To evaluate the diagnostic performance and reliability of magnetic resonance of arthrography (MRA) in diagnosis of subscapularis (SSC) tendon tears between two reviewers with varying levels of experience, and compare the results with clinical tests. SSC tendons were retrospectively evaluated in a total of 272 patients with arthroscopic confirmations. A total of 548 shoulder MRAs were evaluated by two musculoskeletal radiologists, and SSC tendon pathologies were classified into three groups: intact tendon (n = 149), partial-thickness tear (n = 92), or full-thickness tear (n = 31). Diagnostic performance was determined using arthroscopy as gold standard and compared with results of four clinical tests. Intra- and inter-observer reliabilities of two reviewers were evaluated using kappa statistics. For full-thickness tears, mean values of sensitivity, specificity, and accuracy of reviewer 1/reviewer 2 were 71.0%/87.1%, 97.3%/98.3%, and 94.4%/95.5%, respectively. For partial-thickness tears, mean values of sensitivity, specificity, and accuracy were 72.8%/73.4%, 78.3%/81.2%, and 76.5%/78.5%, respectively. Intra- and inter-observer reliabilities for both reviewers were good to very good (k= 0.85/0.93, p < 0.001; k = 0.74-0.89, p < 0.001). For all clinical tests, while specificity was very high, sensitivity was very low and the overall accuracy was also low. MRA showed high diagnostic performance for the diagnosis of SSC tendon tears, especially full-thickness tears, with good inter- and intra-observer reliabilities, regardless of the level of experience of the reviewer.

Artificial intelligence for detection of periapical lesions on intraoral radiographs: Comparison between convolutional neural networks and human observers

The aim of this study was to compare the diagnostic performance of convolutional neural networks (CNNs) with the performance of human observers for the detection of simulated periapical lesions on periapical radiographs. Ten sockets were prepared in bovine ribs. Periapical defects of 3 sizes were sequentially created. Periapical radiographs were acquired of each socket with no lesion and with each lesion size with a photostimulable storage phosphor system. Radiographs were evaluated with no filter and with 6 image filter settings. A CNN architecture was set up using Keras-TensorFlow. Separate CNNs were evaluated for randomly sampled training/validation data and for data split up by socket (5-fold cross-validation) and filter (7-fold cross-validation). CNN performance on validation data was compared with that of 3 oral radiologists for sensitivity, specificity, and area under the receiver operating characteristic curve (ROC-AUC). Using random sampling, the CNN showed perfect accuracy for the validation data. When data were split up by socket, the mean sensitivity, specificity, and ROC-AUC values were 0.79, 0.88, and 0.86, respectively; when split up by filter, they were 0.87, 0.98, and 0.93, respectively. For radiologists, the values were 0.58, 0.83, and 0.75, respectively. CNNs show promise in periapical lesion detection. The pretrained CNN model yielded in this study can be used for further training on larger samples and/or clinical radiographs.

Breast Tumor Classification in Ultrasound Images Using Combined Deep and Handcrafted Features

This study aims to enable effective breast ultrasound image classification by combining deep features with conventional handcrafted features to classify the tumors. In particular, the deep features are extracted from a pre-trained convolutional neural network model, namely the VGG19 model, at six different extraction levels. The deep features extracted at each level are analyzed using a features selection algorithm to identify the deep feature combination that achieves the highest classification performance. Furthermore, the extracted deep features are combined with handcrafted texture and morphological features and processed using features selection to investigate the possibility of improving the classification performance. The cross-validation analysis, which is performed using 380 breast ultrasound images, shows that the best combination of deep features is obtained using a feature set, denoted by features that include convolution features extracted from all convolution blocks of the VGG19 model. In particular, the features achieved mean accuracy, sensitivity, and specificity values of , , and , respectively. The analysis also shows that the performance of the features degrades substantially when the features selection algorithm is not applied. The classification performance of the features is improved by combining these features with handcrafted morphological features to achieve mean accuracy, sensitivity, and specificity values of , , and , respectively. Furthermore, the cross-validation analysis demonstrates that the features and the combined and morphological features outperform the handcrafted texture and morphological features as well as the fine-tuned VGG19 model. The generalization performance of the features and the combined and morphological features is demonstrated by performing the training using the 380 breast ultrasound images and the testing using another dataset that includes 163 images. The results suggest that the combined and morphological features can achieve effective breast ultrasound image classifications that increase the capability of detecting malignant tumors and reduce the potential of misclassifying benign tumors.

Dermoscopic diagnostic performance of Japanese dermatologists for skin tumors differs by patient origin: A deep learning convolutional neural network closes the gap.

In the dermoscopic diagnosis of skin tumors, it remains unclear whether a deep neural network (DNN) trained with images from fair-skinned-predominant archives is helpful when applied for patients with darker skin. This study compared the performance of 30 Japanese dermatologists with that of a DNN for the dermoscopic diagnosis of International Skin Imaging Collaboration (ISIC) and Shinshu (Japanese only) datasets to classify malignant melanoma, melanocytic nevus, basal cell carcinoma and benign keratosis on the non-volar skin. The DNN was trained using 12254 images from the ISIC set and 594 images from the Shinshu set. The sensitivity for malignancy prediction by the dermatologists was significantly higher for the Shinshu set than for the ISIC set (0.853 [95% confidence interval, 0.820-0.885] vs 0.608 [0.553-0.664], P<0.001). The specificity of the DNN at the dermatologists' mean sensitivity value was 0.962 for the Shinshu set and 1.00 for the ISIC set and significantly higher than that for the human readers (both P<0.001). The dermoscopic diagnostic performance of dermatologists for skin tumors tended to be less accurate for patients of non-local populations, particularly in relation to the dominant skin type. A DNN may help close this gap in the clinical setting.

A novel comparative study for detection of Covid-19 on CT lung images using texture analysis, machine learning, and deep learning methods

The Covid-19 virus outbreak that emerged in China at the end of 2019 caused a huge and devastating effect worldwide. In patients with severe symptoms of the disease, pneumonia develops due to Covid-19 virus. This causes intense involvement and damage in lungs. Although the emergence of the disease occurred a short time ago, many literature studies have been carried out in which these effects of the disease on the lungs were revealed by the help of lung CT imaging. In this study, 1.396 lung CT images in total (386 Covid-19 and 1.010 Non-Covid-19) were subjected to automatic classification. In this study, Convolutional Neural Network (CNN), one of the deep learning methods, was used which suggested automatic classification of CT images of lungs for early diagnosis of Covid-19 disease. In addition, k-Nearest Neighbors (k-NN) and Support Vector Machine (SVM) was used to compare the classification successes of deep learning with machine learning. Within the scope of the study, a 23-layer CNN architecture was designed and used as a classifier. Also, training and testing processes were performed for Alexnet and Mobilenetv2 CNN architectures as well. The classification results were also calculated for the case of increasing the number of images used in training for the first 23-layer CNN architecture by 5, 10, and 20 times using data augmentation methods. To reveal the effect of the change in the number of images in the training and test clusters on the results, two different training and testing processes, 2-fold and 10-fold cross-validation, were performed and the results of the study were calculated. As a result, thanks to these detailed calculations performed within the scope of the study, a comprehensive comparison of the success of the texture analysis method, machine learning, and deep learning methods in Covid-19 classification from CT images was made. The highest mean sensitivity, specificity, accuracy, F-1 score, and AUC values obtained as a result of the study were 0,9197, 0,9891, 0,9473, 0,9058, 0,9888; respectively for 2-fold cross-validation, and they were 0,9404, 0,9901, 0,9599, 0,9284, 0,9903; respectively for 10-fold cross-validation.

Octopus 900 Automated Kinetic Perimetry versus Standard Automated Static Perimetry in Glaucoma Practice

ABSTRACT Purpose: The presence of central visual field loss does not infer the extent of peripheral visual field loss. In advanced stage glaucoma, we evaluated whether automated kinetic perimetry provided additional visual field information to that of central static perimetry. Materials and Methods: We undertook a prospective cross-sectional study of advanced-stage glaucoma defined as stages 3–4. Visual field assessment for right and left eyes was undertaken within one clinic visit using the Octopus 900 G programme and kinetic strategy. Results: We recruited 126 patients (170 eyes). Mean patient age at assessment was 55.86 years (SD 15.15). Mean kinetic reaction time was 1503.96 ms (SD 801.68). Kinetic I4e was plotted in 71% of eyes with an unadjusted area of 2513.68 degrees2 (SD 2397.91) and mean isopter radius of 23.16 degrees (SD 13.07). Kinetic I2e was plotted in 53.5% of eyes with an unadjusted area of 627.07 degrees2 (SD 1291.94) and mean isopter radius of 7.47 degrees (SD 10.59). Increased reaction time was associated with a poorer visual field (p = .001). Mean sensitivity, mean deviation and standard loss variance values on static perimetry were higher in patients who had a defined kinetic field boundary than in patients with no kinetic response to I4e stimulus (p = .0001). However, this corresponded to only small-to-medium correlation between static fields and existent kinetic fields: the presence of poor static fields did not always infer a poor kinetic visual field as poor static fields could also have good kinetic visual fields. Conclusions: Although we confirmed a lack of agreement and only a small to medium correlation between the extents of central versus peripheral visual field loss, automated kinetic perimetry did provide additional peripheral (outside the static 30-degree central field) visual field information which was clinically useful in the presence of non-informative severely defected central visual fields.

Effect of Caffeine on Contrast Sensitivity among Young Adults of Abia State University

Introduction: Caffeine is the most widely consumed psychoactive substance. Unlike many other psychoactive substances, it is legal and unregulated in the world. Several experts suggest that moderate use of caffeine (300 mg or equivalent to 3 cups or less to regular size cups of coffee per day) is safe and so not likely to cause health issues. Intake of caffeine has been suggested to affect the visual system. This study examined the effect of consumption of caffeine on contrast sensitivity of the eye among young adults of Abia State University students within the age range of 17-30 yrs.&#x0D; Methods: This study employed a cross-sectional design and was carried out within the confinement of Abia State University Eye Clinic. A total of 100 young adults consisting of an equal number of males and females were measured using the Rabin’s contrast sensitivity at baseline, 30 mins until 120 mins. One way analysis of variance (ANOVA) and the Chi-square statistical analysis was used to analyze statistically the research hypotheses at 95% level of significance.&#x0D; Results: The findings of this study reveal that mean contrast sensitivity values for 16 to less than 20 years old were 1.70 at baseline and increased to 1.79, 1.89, 1.93 except at 120 minutes post-ingestion of caffeine. In males and females, contrast sensitivity increased by 1.57%, 3.7%, 5.3% and 10.7%, 20.7% 23.3% respectively. At 95% confidence level, caffeine ingestion had a significant effect on contrast sensitivity for both males (4.79) and females (5.8).&#x0D; Conclusion: This study depicts that the ingestion of caffeine affects or increases contrast sensitivity in both males and females although this effect is short-lived.